The nature of purpose relatives: crosslinguistic and acquisition evidence
Inês Duarte, Ana Lúcia Santos & Nélia Alexandre
Universidade de Lisboa
A few studies show the structural parallel between purpose relatives and wh-/that-relatives in
English (Jones 1991, Beavers & Bender, 2004, Bhatt 2006). However, to our knowledge, work
discussing the nature of similar constructions in other languages is scarce (Schmidtke-Bode 2009). This
paper addresses the status of purpose relatives in (European and Brazilian) Portuguese and Cape
Verdean Creole (CVC).
Based on constituency tests (1) and on parasitic gap effects (2), we show that a subset of clauses
introduced by para ‘to’ in Portuguese and pa ‘to’ in CVC are purpose relatives with subject, object and
oblique gaps. CVC data is particularly clear concerning the nature of this structure: spelled-out traces,
known to occur in CVC PP ki ‘that’-relatives derived by Move (Alexandre, 2009), are available in this
type of relatives (3). Apart from para / pa headed relatives (4), we will argue that we can also find free
relatives (5).
(1)
(2)
(3)
a. Ele tem
um livro para
ler.
he has a book
to to-read
b. É
[um livro para ler]
que
(it) is [a book to read] that
É
(it) is
um livro
a book
para
to
(EP, BP)
ele tem.
he has
ler
sem
to-read without
estragar
to-spoil
Bu tene uns faka mau pa korta karne ku-el/*es.
(CVC)
2SG have a.PL knife bad to to-cut meat with-3SG/3PL
(4)
Comprei
I-bought
uma
estante para
pôr livros.
a bookshelf to to-put books
(5)
a. Essa
that
estante
bookshelf
é para
is to
b. Banhera e pa da
mininu
Bathtub is to to-give child
pôr
livros de arte.
to-put books of art
banhu.
bath
(EP, BP)
(CVC)
Following Kayne (1994) and the refinement in Bianchi (1999), we assume that that-relatives
involve raising of the head. However, we will argue that this type of movement is absent in purpose
relatives. Purpose relatives involve raising of a null operator to Spec,CP, whose range is fixed by the
closest antecedent – the closest antecedent being the head of the relative, in headed purpose relatives.
This analysis will allow us to explain two types of facts, one concerning pa relatives in CVC and
another one concerning the acquisition of Portuguese.
As shown by Alexandre (2009), CVC does not show PP movement in relatives and it also
precludes preposition stranding. The first fact excludes PP gaps in ki-relatives (6b) and the second fact
justifies the occurrence of spelled-out traces and resumptive pronouns in PP ki-relatives (6a). In
contrast, CVC PP pa purpose relatives may present gaps (7): this fact is explained if pa relatives do not
involve raising of the relative head; instead they involve raising of a null operator to Spec, CP.
(6) a. Djon atxa
kes
mudjeris ki Zé papia
Djon find(PFV) DET women
that Zé talk(PFV)
b. *Djon atxa kes mudjeris (ku) ki Zé papia.
ku-el/es.
with-3SG/3PL
(7)
N
ka odja
1SG NEG see(PFV)
un txabi pa-N
abri porta.
a key
to-1SG open door
On the other hand, Alexandre (2009) also shows that PP ki-relatives display resumptive
pronouns when they do not involve movement to Spec, CP, whereas they involve a spelled-out trace
when the head of the relative is subextracted from the PP. The fact that we can also have spelled-out
traces in pa relatives (but not resumptive pronouns – see (3)) follows from our analysis: (i) pa relatives
always involve Move and (ii) since the moved element is a null operator, which arguably has a reduced
set of formal features, its tail may be pronounced as a spelled-out trace (Alexandre, 2009).
Our analysis also accounts for the early acquisition of non-subject para-relatives in Portuguese.
The acquisition of relatives is known to be problematic (Sheldon 1974, Tavakolian 1981, Hamburger &
Crain 1982 and subsequent work), especially in what concerns object relative clauses (see discussion in
Friedmann & Novogrodsky, 2004). A recent proposal by Friedmann, Belletti & Rizzi (2009) suggests
that the difficulty with object relatives is an effect of an extension of Relativized Minimality operating
in child language (3;7-5;0). This effect is produced by an intervening subject with a subset of features
identical to those of the chain headed by the “antecedent” in headed relative clauses.
Using a corpus representative of the acquisition of three Portuguese monolingual children
(1;6.6–3;11.12, 1;6.18–2;9.7, 1;5.9–2;7.24; MLUw 1.3-4.7), we coded all cases of that- and parapurpose relatives. The first set of results concerns the precocity of purpose relatives, which in the
period between 2;0 and 2;6 are more frequent than that-relatives (33 purpose relatives out of 43 relative
clauses) and include both headed (8) and free relatives (9). The second set of results concerns the site
where the relativized constituent is merged: whilst that-relatives between 1;5 and 3;11 mostly present
relativized subjects, purpose relatives are object or oblique relatives ((8) and (9)). The preliminary
analysis of data of one child acquiring BP (2;4.11 – 2;10.29; MLUw 1.5-2.9) confirms the precocity of
para-purpose relatives in production (10).
We extend the proposal by Friedmann, Belletti & Rizzi and suggest that non-subject purpose
relatives are not problematic for children since they do not create the type of intervention effects that
have been reported in that-relatives: given our analysis of purpose relatives in Portuguese and CVC,
they always involve a chain with no lexical or phonological features (in Portuguese), contrary to thatrelatives.
(8)
child: ce(re)ais # pa(ra) eu comer.
Cereals to
I to-eat
‘Cereals for me to eat.’
(9)
a. adult:
olha [/] # olha # e
look
look
and
child: é
pa(ra) pôr
a
(it) is to
to-put the
b. adult:
child:
(10)
(2 ;5.24)
isto
é para o quê?
what is this for?
ma(n)ga.
sleeve
que
é isso?
what is this?
é
pa(ra) a boneca mo(r)der.
(it) is to
the doll
to-bite
‘it is for the doll to bite’
child: esse
that
chapéu #
hat
p(r)a [: para]
to to-change
(2;2;1)
(2;3;8)
t(r)oca(r) !
(2;5.21)
Download

The nature of purpose relatives: crosslinguistic and acquisition